Skip to main content
AI-native backend · Mini Program / Web / App, fully covered

AI writes the code,
CloudBase runs the backend

An all-in-one backend platform built for AI agents. Built-in database, storage, authentication, cloud functions, and Cloud Run, deeply integrated with the WeChat ecosystem. Codex, Claude Code, WorkBuddy, and other AI agents connect with one command — from writing code to shipping in production.

Leading AI coding & agent platforms use CloudBase as their backend foundation

TusiTusiCodeBuddy GenieCodeBuddy GenieWorkBuddyWorkBuddy
The Last Mile

Why do you always get stuckright after the code is done?

It's not that you can't write the code — environment setup, resources, deployment, and ops together are enough to drag a perfect demo into endless delay. You've probably run into at least one of these:

Environments built from scratch, every time

Database, storage, auth... AI finishes the code, then gets stuck because the environment isn't set up yet.

Deployment needs a dozen skills

SSH, domains, certificates, servers — agents don't know any of this, and errors hit the moment they try.

Re-integrating an SDK for every platform

One SDK for Mini Program, one for Web, one for Node — every new framework means retracing the same pitfalls.

WeChat capabilities don't plug into AI

Login, payments, messaging — these native WeChat capabilities are hard to fit into an AI workflow.

Launch is only the beginning

Scaling, monitoring, alerts, rollbacks — ops is a whole new pile of work, and the demo never quite makes it past rehearsal.

One backend, covering 12dev frameworks & runtimes

WeChat Mini ProgramMini GameReactVueNode.jsPythonGoJavaFlutterSwiftUniAppCocosWeChat Mini ProgramMini GameReactVueNode.jsPythonGoJavaFlutterSwiftUniAppCocos
The Fix · One Serverless Backend

Fold the "last mile"
into one platform

Database, storage, auth, functions, hosting, AI — all ready out of the box, sharing one identity and one SDK. AI provisions it in one step; you just write the business logic.

Database · Dual Engine

PostgreSQL relational + document DB, SQL, real-time subscriptions (watch), row-level security — as smooth as Supabase.
todos · SQL + watch + RLS
SELECT * FROM todos WHERE _openid = $1;
titledonePermission
Hello CloudBasefalseRLS ✓
Ship v2trueRLS ✓

Authentication

WeChat auth-free, anonymous, phone number, and custom login.
auth
user_8f2a
OPENID ✓ verified
Session ✓ wx-server-sdk

Cloud Storage

Files / media / image processing, served directly via CDN.
storage
📄 doc.pdf🖼 img.png
🎞 video.mp4📦 pkg.zip
upload.mp4 ▲ 78%

Cloud Functions & Cloud Run

Event / HTTP / edge triggers with auto-scaling; use Cloud Run for containerized services.
functions
trigger: HTTPonOrder()DB.write
cron: dailyaggregate()notify
~ cold start 32ms · auto-scales 0→N

Static Hosting

One-click upload, HTTPS + global CDN.
hosting
HTTPS ✓
200+ CDN nodes
~ ~ edge latency enabled ✓

AI Capabilities

Vector store, embeddings, generation, and retrieval in one.
ai / rag
user ▸ LLM
· · · · · · · · · · vector scatter · · · · · · · · · ·
emb: v1536retrieve ✓

Same task, CloudBase cuts out most of the detours for AI

Source: docs.cloudbase.net/solutions/vibe-coding-platform/vm-vs-cloudbase-comparison (model: glm-5.0-ioa)

3.8×
Faster completion
990s → 260s
52%
Token savings
Leaner context
5.4×
Less ops overhead
76% → 14%
2.2×
Fewer tool calls
79 → 36 calls

Traditional VMs fail 50% of the time; CloudBase has zero SSH exposure risk, with HTTPS + platform signature verification.

AI Dev Toolkit

Your AI IDE can already run the backend

MCP · Skills · Plugin, one setup. WorkBuddy, ZCode, Codex, Cursor, and other mainstream AI coding environments work out of the box.

View all setup options →
Just tell the AI

Tell the AI to use CloudBase like this

Open AI Agent
1Copy the AI Prompt

Paste into your AI IDE

Set up CloudBase for me: do the following.
1. Open https://docs.cloudbase.net/skill.md and follow the instructions.
2. Tell me when it's done, and suggest the best next step.
2Develop with Skills

Say this to the AI

Use CloudBase Skills to build a full-stack web app and deploy it
Developer Tools

From CLI to SDK, the full chain, connected

Whether you prefer CLI, MCP, Skills, or SDKs in any language, CloudBase has an entry point for you — AI and human developers share the same interface. More in the Developer Hub.

AGENT SDK

Build your own agent — kernel included out of the box.

Multi-turn conversations, session persistence, and tool calling are all built in, compatible with the Claude Agent SDK ecosystem.

CloudBase Agent SDK

Official Tencent Cloud SDK · one line with envId connects to CloudBase resources; sessions and files live in your own environment.

HITL approvalMCP tool integrationSandboxed executionMultimodal attachments
$npm i @cloudbase/open-agent-kernel@betaCopy

Leading agent platforms,
already running on CloudBase

The cloud is becoming the software production line for agents · Read the WAIC talk

WeChat Ecosystem · Official Backend Foundation

Native WeChat capabilities, callable auth-free

Jointly launched by Tencent Cloud and the WeChat Mini Program team. Mini Programs use cloud calls to access WeChat's open capabilities auth-free, and cloud functions get OPENID automatically via wx-server-sdk. Develop once, run everywhere. See WeChat Mini Program AI Solution · Authentication.

01

Auth-free cloud calls

Mini Programs call WeChat login, payment, messaging, and other open capabilities directly — no need to build your own auth service or manage token flows.

02

OPENID obtained automatically

Cloud functions get user identity automatically via wx-server-sdk — secure and trusted, eliminating the frontend parameter-trust problem.

03

Develop once, run everywhere

Mini Program / Official Account / Web / mobile share one backend and one dataset, with native SSE / WebSocket real-time support.

Join the community

Discuss CloudBase and AI app development with developers — feedback and feature requests live here.

Hands-on video tutorials by community creators, updated regularly.

Open Source

The whole toolchain is open source

CloudBase's SDKs, CLI, MCP, AI Toolkit, and app templates are all open source on GitHub and CNB. Going global has just begun — come be one of the first contributors.

Let AI take the code,and actually ship it

Install CloudBase Skills, tell AI one sentence, and let the platform handle environment, configuration, and deployment.